ArhGAP9 and ArhGAP12 are highly similar RhoGAP proteins which possess the SH3, WW, PH and RhoGAP domains. We have cloned the full-length mouse ArhGAP9 cDNA (GeneBank accession no. DQ498128) and found that although it shares 64% sequence identity with human ArhGAP9, it lacks the WW domain (Fig. 1a). According to the SMART domain prediction tool 32, bovine, canine and primate ArhGAP9 contain the SH3, WW, PH and RhoGAP domains, murine and rat ArhGAP9 contain the SH3, PH and RhoGAP domains and lack the WW domain. It is possible that genomic deletion resulting in the loss of the WW domain in ArhGAP9 has occurred rodents. To understand the signaling function of the WW domain in ArhGAP9 of organisms that retain this domain, we screened for interacting proteins for the WW domain of human ArhGAP9. The GST fusion protein of the WW domain of human ArhGAP9 immobilized on glutathione Sepharose beads was used as a bait to isolate interacting proteins from rat brain lysate using proteomics (Fig. 1b). Several specific binding proteins were identified, including the MAP kinase Erk2 from bands f and g that generated 15 unique peptides covering 40% of the protein sequence of Mitogen-activated protein kinase 1 (MAPK1, also Erk2). The other bands corresponded to other proteins were characterized separately from this report.

Erk2 binding to the WW domain of human ArhGAP9 was confirmed by in vitro pulldown assays (Fig. 2a). Flag-tagged Erk2 was expressed in 293T cells and the lysates prepared were incubated with GST fusion protein of the WW domain of ArhGAP9 immobilized on glutathione Sepharose beads (ArhGAP9-WW-GST). The binding of two closely related MAP kinases, p38α and Jnk1, was also tested. Consistent with the proteomics data, Erk2 was precipitated by ArhGAP9-WW-GST [Fig. 2a(i)]. While p38α also showed binding albeit to a lesser extent than Erk2, the binding of Jnk1 was not detectable in the pulldown assay. Full-length ArhGAP9 was also shown to be able to bind to immobilized recombinant Erk2 or p38α (Fig. 2b). Using coimmunoprecipitation experiments, the binding of Erk2 or p38α to ArhGAP9 in vivo was confirmed [Fig. 2c(i)]. Given that Erk2 and p38α showed no binding to the two WW domains of ArhGAP12 (ArhGAP12-WW1 and ArhGAP12-WW2) or the first WW domain of Nedd4 (Nedd4-WW1) [Fig. 2d(i)], the binding of Erk2 and p38α to the WW domain of ArhGAP9 was specific.

As mouse ArhGAP9 does not contain a WW domain, we tested whether the protein would be able to interact with the MAP kinases Erk2 and p38α like the human ortholog. A GST fusion of an N-terminal fragment of the mouse ArhGAP9 protein (mArhGAP9-N), comprising of amino acids 1–350 which included the SH3 and PH domain and the intervening sequence between these two domains was found to be unable to bind to Erk2 and p38α, when compared to human ArhGAP9 WW domain [Fig. 2e(ii)]. Using coimmunoprecipitation experiments, it was confirmed that full-length mouse ArhGAP9 could not interact with Erk2 in vivo (Fig. 2f). Taken together, we conclude that human ArhGAP9 interacted with Erk2 and p38α specifically with its WW domain and the mouse ortholog which lacked a WW domain was unable to bind to MAP kinases.

Sequence alignment of human ArhGAP9-WW domain with all the known WW domains showed that the C-terminal end of human ArhGAP9-WW domain contained a unique basic di-Arginine motif (R246 and R247) that are not present in all other WW domains compared (data not shown). The alignment of the WW domains of human ArhGAP12 and ArhGAP9 showed that in the first and second WW domains of ArhGAP12 which did not bind Erk2 and p38α, the amino acids in alignment with R247 are hydrophobic residues W and Y, respectively (Fig. 3a). Such short basic motifs, termed Docking (D) domains, had been found to be present in many MAP kinase docking proteins that bind to the conserved docking (CD) domains of MAP kinases 33343536. The CD domains of Erk2, p38 and Jnk consist of conserved acidic residues which form electrostatic interactions with the basic residues of the target docking proteins 33343536. Therefore we examined the possibility that the interaction between the complementarily charged residues in the CD domains of Erk2 or p38α and WW domain of ArhGAP9 could be the mechanism which enhanced or mediated the binding of these MAP kinases and ArhGAP9. We investigated whether the basic residues in the WW domain of human ArhGAP9 would mediate the binding to Erk2 and p38α through interaction with the acidic residues in the CD domain, as in the case of other MAP kinase docking proteins. R246 and R247 in the WW domain of human ArhGAP9 were mutated to Alanine (ArhGAP9-WW-RR) and the binding of Erk2, p38α or Jnk1 was tested. As shown in Fig. 3b(i), for both Erk2 and p38α, the binding to the WW domain with R246, 247A mutation was significantly reduced compared to binding to wild type ArhGAP9 WW domain, confirming that these basic residues played a role in MAP kinase interaction. Single mutation of R246A or R247A was also defective in binding to Erk2 (Fig. 3c). The binding of full-length R246, 247A (RR) mutant of ArhGAP9 to Erk2 or p38α was significantly reduced relative to wild type ArhGAP9 in vivo using coimmunoprecipitation experiments (Fig. 3d). Therefore, R246 and R247 in the WW domain of ArhGAP9 are important in mediating the binding to Erk and p38 MAP kinases. Further binding analysis indicated that residues in the vicinity of this di-Arginine motif, namely, K243, P244 and P245 with the WW domain also contributed to binding of ArhGAP9 with MAP kinases (Fig. 3e).

To examine whether the CD domains in the MAP kinases Erk2 and p38α were the regions that interacted with the WW domain of ArhGAP9, GST fusions of a C-terminal fragment of Erk2 containing the CD domain (Erk2-CD-GST, residues 300–358) and a deleted version of Erk2 lacking the CD domain (Erk2-ΔCD-GST, residues 1–300) were constructed. The binding of wild type or R246, 247A mutant of ArhGAP9 with immobilized Erk2-CD-GST and Erk2-ΔCD-GST were tested [Fig. 4a(i)]. Wild type ArhGAP9 but not the R246, 247A mutant, specifically precipitated with Erk2-CD-GST, indicating that the CD domain was sufficient to mediate Erk2 binding to ArhGAP9. Consistent with this notion, Erk2-ΔCD-GST did not bind wild type ArhGAP9 [Fig. 4a(ii)].

To further determine whether the conserved acidic residues [Fig. 4b(i)] in the CD domains of the MAP kinases were important in mediating binding to the WW domain of ArhGAP9, point mutants of Erk2 and p38α as indicated in Fig. 4b(ii) were generated and tested for their binding to ArhGAP9. As shown in Fig. 4b(ii), the mutations D316A in Erk2 and D312A, D315A, and E316A in p38α caused a significant reduction in interaction with ArhGAP9-WW domain. These results clearly indicated that the acidic residues in the CD domains of Erk2 and p38α were important for their interactions with ArhGAP9. Other point mutations of Erk2 and p38α including amino acid substitutions in the activation loop of the kinase domain did not affect the binding (Fig. 4c). Further evidence confirming that the WW domain of ArhGAP9 interacted directly with the CD domains of Erk2 and p38 but not Jnk1 was obtained by Far-UV circular dichroism spectroscopy (Fig. 4d). The WW domain is a small three-stranded β-sheet stabilized by the stacking of several conserved aromatic and proline residues 37383940. Crystal and solution structures of WW domains had shown that WW domains undergo conformational changes upon binding to its peptide ligands 192041. We applied Far-UV circular dichroism spectroscopy to investigate the conformational changes upon binding of Erk2, p38α or Jnk1 peptides corresponding to the common docking domains to the WW domain of ArhGAP9. The peptides containing 14 amino acids in common docking domains of Erk2, p38α and Jnk1 were synthesized [sequences shown in Experimental procedures and in Fig. 4b(i)]. To analyze the changes in the structure of WW domain upon binding of corresponding peptides, the differences in the Far-UV circular dichroism spectrum of ArhGAP9 WW domain solutions in the presence or absence of the peptides were monitored. The changes in secondary structure contents were analyzed using CDNN program, version 2.1 42. The analysis showed that the recombinant ArhGAP9 WW domain folded properly and possessed mostly β-sheets and random coils which are similar to the known WW domains [Fig. 4d(i)]. In the presence of the Erk2 peptide at equimolar concentration, ArhGAP9 WW domain showed significant difference in the circular dichroism spectra particularly from 190–205 nm [Fig. 4d(iv)]. In the presence of p38α peptide, ArhGAP9 WW domain also showed significant difference in the circular dichroism spectra in the same region [Fig. 4d(iii)]. In contrast, the addition of Jnk1 peptide did not alter the circular dichroism spectra significantly although there were notable changes at 220–235 nm [Fig. 4d(ii)]. These conformational changes monitored by Far-UV circular dichroism were in agreement with the other known crystal and solution structures of WW proteins with their peptide targets. This result further supported the conclusion from the in vitro and in vivo experiments that both Erk2 and p38α interacted with ArhGAP9-WW domain but Jnk1 did not. However, the exact conformational changes and mechanism of binding can be obtained only by solving the crystal or solution structures of ArhGAP9-WW domain in complex with its binding partners. The data described above strongly indicated that the binding between ArhGAP9 and the MAP kinases Erk and p38 is specific and is mediated by the WW and CD domains of ArhGAP9 and the MAP kinases, respectively.

To gain insights into the structural basis of specificity of the WW domain of ArhGAP9 binding to Erk2, p38α and Jnk1, we compared the three-dimensional structures of CD domains of the three proteins. It could be seen from the comparison that although the overall structures of Erk2 (PDB id: 1ERK), p38α (PDB id: 1P38) and Jnk1 (PDB id: 1JNK) are very similar, the CD domain of p38α shared greater homology than that of Jnk1 to Erk2. Furthermore while Erk2 and p38α are significantly super-imposable with RMSD of 0.68 Å [Fig. 4e(i)], Erk2 and Jnk1 CD domains are relatively distinct with RMSD of 2.18 Å [Fig. 4e(ii)]. The high sequence and structural similarity between Erk2 and p38α provided some insights into the structural basis for the specificity on the WW domain of ArhGAP9 for Erk2 and p38α but not Jnk1, as observed in the binding studies in vitro and in vivo.

The CD domains of MAP kinases had been established as the binding sites for their upstream activating kinases, namely MEK1 and 2 for Erk1 and 2, and MKK3, 4 and 6 for p38α. 33353643. Some other regulators, effectors as well as MAP kinase substrates have also been established to dock on the CD domains. These docking proteins all contain positively charged residues that form the docking (D) domain for binding with the CD domain of MAP kinases 23252627. From our data that the binding of ArhGAP9 to Erk2 and p38α required a di-Arginine motif in the WW domain and conserved acidic residues in the CD domain, it was possible that the binding of D domain containing proteins such as MEK2 and MKK6 may compete with the binding of ArhGAP9 to Erk2 and p38α, respectively. We therefore compared the binding of p38α to ArhGAP9-WW domain in the absence or presence of overexpressed MKK6 [Fig. 5a(i)]. The amount of p38α that precipitated with the WW domain ArhGAP9 was significantly reduced in the presence of MKK6, indicating that it was possible that the binding of MKK6 to p38α could inhibit the binding of the WW domain ArhGAP9. Consistent with the above in vitro data, the binding of full-length ArhGAP9 to p38α in vivo was reduced in the presence of MKK6 (Fig. 5b). We had also observed that a mutant of MKK6 with the basic residues in the D domain responsible for binding to the CD domain of p38α deleted failed to bind p38α and therefore to inhibit p38α binding to the WW domain ArhGAP9 (data not shown). Similarly, with the cotransfection of MEK2, the amount of ArhGAP9 that co-immunoprecipitated with Erk2 was significantly reduced (Fig. 5c). Taken together, these results indicated that upstream activating kinases of p38α and Erk2, namely MKK6 and MEK2, respectively were able to disrupt the binding of the MAP kinases to the WW domain of ArhGAP9, most likely due to their docking onto the CD domains and competitively preventing the binding of ArhGAP9.

Given that both ArhGAP9 and the MAP kinases possess catalytic activities, we addressed the question of whether their interaction would influence their activities. Furukawa et al 717 as well as our data unpublished showed that ArhGAP9 functioned as an active GTPase activating protein (GAP) towards cdc42 and Rac1 but not RhoA. To determine whether MAP kinase binding had any effect on the RhoGAP activity of ArhGAP9, we compared the levels of active cdc42 in the lysates of cells transfected with ArhGAP9 in the presence or absence of Erk2 or p38α using the Pak-binding domain (PBD) GST method (Fig. 5d). Through protein sequence alignment of the RhoGAP domain of human ArhGAP9 and other known RhoGAPs whose activities and structures have been determined, we identified R578 in human ArhGAP9 as the 'Arginine finger' residue essential for GAP activity. We mutated R578 was to lysine and confirmed that ArhGAP9 (R578K) was catalytically inactive by in vitro GAP assays (data not shown). Using the GAP-inactive mutant of ArhGAP9 (R578K) as a negative control for the assay, there was clearly no significant effect in the activity with or without the presence of Erk2 or p38α.

We next investigated whether the binding of ArhGAP9 to the Erk2 and p38α had an effect on the MAP kinase activities. 293T cells were transfected with p38α-Flag alone, or p38α-Flag with EGFR to induce the activation of p38α, or p38α with EGFR as well as ArhGAP9, either wild type or the W242K, R246, 247A (WRR) triple mutant that was defective in binding to Erk and p38, as indicated in Fig. 5e. As shown by western blotting with α-phospho-p38 in Fig. 5e(v), p38α transfected alone showed a slight detectable level of basal activation which increased significantly with the coexpression of EGFR. With the expression of wild type ArhGAP9, a reduction of p38α activation induced by EGFR was observed, indicating that ArhGAP9 could suppress the activation of p38α downstream of EGFR. With coexpression of the ArhGAP9 (WRR) mutant, there was a slighter suppression of p38α activation by EGFR indicating that this mutant was able to partially restore p38α activation by EGFR inhibited by wild type ArhGAP9. This result strongly suggested that the binding of ArhGAP9 to MAP kinases prevented the activation of the latter by their upstream kinases.

We further confirmed that this inhibitory effect was not attributed to the RhoGAP activity but to the WW domain of ArhGAP9. 293T cells were transfected with p38α-Flag alone, or p38α-Flag with EGFR, or p38α with EGFR as well as ArhGAP9, either wild type, the GAP-inactive mutant (R578K) or the WW domain mutants (W219K or W242K), as indicated in Fig. 5f. As shown in Fig. 5f(iv), the suppressive effect of wild type ArhGAP9 on EGFR-induced p38 activation was not affected by the R578K mutation, but was partially reversed by the W219K or W242K mutation.

Taken together, the above data demonstrated that (a) the binding of ArhGAP9 to MAP kinases through its WW domain caused the inhibition of MAP kinase activation by upstream signals and (b) this interaction did not have a significant effect on the RhoGAP activity of ArhGAP9. (c) In addition, the inhibitory effect of ArhGAP9 on MAP kinase activation was independent of its RhoGAP activity.

Rho GTPase and MAP kinase signaling have both been implicated in the control of the actin cytoskeleton. Specifically it was reported that activation of the MAP kinases Erk1, 2 and 5 pathways caused disruption of the actin cytoskeleton 44. To investigate if the interaction of ArhGAP9 and MAP kinases had an effect on the actin cytoskeleton, ArhGAP9 (wild type or R246, 247A mutant) and GFP-actin were heterologously expressed in Swiss 3T3 cells by microinjection of the respective cDNAs, followed by live imaging of the actin structure. As shown in Fig. 6a, in the presence of wild type ArhGAP9, the cells possessed distinct bundles of actin stress fibres as in the control cells where GFP-actin alone was expressed. However, in the presence of the R246, 247A ArhGAP9 mutant, the stress fibres dissolved and showed a diffuse cytosolic pattern of actin distribution. These results supported the hypothesis that binding of wild type ArhGAP9 to MAP kinases sequestered them in the inactive forms, preventing the activation of the MAP kinases and therefore allowing the formation and maintenance of actin stress fibres in fibroblasts. Conversely, the R246, 247A mutant being unable to bind the MAP kinases, would potentially allow activation of the MAP kinases to result in the disassembly of the actin stress fibres.

α-Flag (M2) and α-phospho-p38 were from Sigma, α-phospho-Erk was from Cell Signaling Technology, α-HA (12CA5) was from Boerhinger, α-ArhGAP9 was generated by standard immunization procedure with full-length human ArhGAP9-GST as the immunogen. α-phosphotyrosine conjugated to horse radish peroxidase (PY20-HRP) was from Transduction Laboratories, α-EGFR was from Santa Cruz Biotechnology.

293T cells were cultured in Dulbecco's minimal essential medium (DMEM) containing 4.5 g/l D-glucose, 10% (v/v) fetal bovine serum (HyClone), 10 mM L-glutamine, and 100 μg each of penicillin and streptomycin/ml from Sigma. Swiss 3T3 were cultured in DMEM containing 4.5 g/l D-glucose, 10% (v/v) Cosmic calf serum (HyClone), 10 mM L-glutamine, and 100 μg each of penicillin and streptomycin/ml.

Human ArhGAP9 (wild type and mutants) and EGFR coding cDNAs were cloned into the mammalian expression vector pRK5. Erk2, p38, Jnk1 cDNAs were cloned in pxJ40-Flag vector with an N-terminal Flag tag. HA-tagged activated MEK2(S222D, S226D) was cloned in pUSEamp vector. Activated MKK6(S207E, T211E) was cloned in pxJ40-HA vector. Plasmid encoding Pak-binding domain (PBD)-GST was obtained from Dr. E Manser (CMM, A-STAR, Singapore). Cdc42-myc was cloned in pcDNA3. For transient expression in mammalian cells, Lipofectamine (Gibco BRL) was used for transfection, following the protocol recommended by the manufacturer. Point mutations were introduced with the Quikchange site-directed mutagenesis kit (Stratagene) and the mutations were confirmed by DNA sequencing.

For expression of glutathione S-transferase (GST) fusion proteins, cDNA sequences were cloned into the pGEX4T1 vector (Pharmacia). The GST fusion proteins were expressed in E. coli BL21 (DE3) cells and purified using glutathione sepharose 4B column chromatography (Amersham). For Far-UV CD spectroscopy, the GST tag was cleaved with the thrombin protease by overnight on-column digestion at 4°C. The resulting protein preparation was then further purified by using ion-exchange Mono Q Sepharose column (Amersham) which had been pre-equilibrated with buffer A (20 mM Tris-HCl pH 7.5). The protein was then eluted from the column with a linear gradient to buffer B (20 mM Tris-HCl pH 7.5, 1 M NaCl). The protein was further purified using Superdex-75 gel filtration column chromatography in Biologic Duoflow FPLC system.

Protein complexes of the WW domain of human ArhGAP9 isolated from rat brain lysate were eluted and resolved by 10% SDS-PAGE, followed by detection of bound proteins by staining with Colloidal Coomassie Blue (Pierce). Specific bands were excised and subjected to in-gel reduction, S-alkylation and trypsin hydrolysis. Liquid-chromatography tandem mass spectrometry (LC-MS/MS) analysis of the peptides was performed on a Finnigan LCQ Deca ion trap mass spectrometer (Thermo Finnigan) fitted with a Nanospray source (MDS Proteomics). Chromatographic separation was conducted using a Famos autosampler and an Ultimate gradient system (LC Packings) over Zorbax SB-C18 reverse phase resin (Agilent) packed into 75 μm PicoFrit columns (New Objective). Protein identifications were made using the search engines Mascot (Matrix Sciences) and Sonar (ProteoMetrics).

Lysates for binding assays and immuno-precipitation were prepared by lysis of cells in the cell lysis buffer [20 mM HEPES (pH 7.5), 137 mM NaCl, 1% Triton X-100, 10% glycerol, 1.5 mM MgCl₂, 1 mM EGTA, 0.1 mM Na₃VO₄ and Complete Protease inhibitors (Boehringer)], followed by centrifugation 13,000 g for 15 min at 4°C and collecting the supernatant. For in vitro pulldown assays, GST fusion proteins immobilized on glutathione beads were incubated with the lysates for 1 hour at 4°C with rotation, followed by washing of the beads with specifically bound proteins with cell lysis buffer for 5 times, each time for 5 min with rotation at 4°C. The protein complexes were then eluted with Laemmli Buffer and resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by blotting onto PVDF membranes and detection with specific antibodies. For immuno-precipitation, antibodies were added to the lysates and incubated for 1 hour at 4°C with rotation, followed by addition of Protein A PLUS G beads (Calbiochem) to capture the immuno-complexes. The immuno-complexes were washed and then eluted with Laemmli Buffer and resolved by SDS-PAGE, followed by western blotting with specific antibodies.

Peptides for Erk2: LEQYYDPSDEPIAE, p38α: FAQYHDPDEPVAD and Jnk1: INVWYDPSEAEAPP were custom synthesized by NeoSystems (Strasbourg, France). CD experiments were performed with a Jasco J-715 spectropolarimeter equipped with a Peltier cell holder and a PTC-348WI temperature controller. Far-UV CD spectra were measured in a 1-mm rectangular quartz cell. The buffer used was 20 mM Tris-HCl (pH 7.5), 500 mM NaCl and 1 mM DTT. Protein concentration of approximately 0.5 mg/ml determined by the Bradford assay method was used for wavelength scans. Wavelength scans were made at a scan rate of 10 nm/min and the average value of 3 scans of the same solution were obtained. Data were collected at 25°C over a wavelength range 190–260 nm with a bandwidth of 1 nm. To determine the conformational changes in WW domain upon peptide binding, the ellipticity of the corresponding peptides at equimolar concentration with the solvent were subtracted from the ellipticity of the WW domain-peptide complex and similarly the solvent spectrum were subtracted from the spectrum of WW domain before analysis. The far UV CD spectra were analyzed by using the secondary structure analysis program CDNN, version 2.1 42.

Swiss 3T3 fibroblasts were plated at 5 × 10⁴ per glass bottom dish and grown overnight at 37°C in DMEM with 10% Cosmic calf serum. DNA was injected at 0.5 μg/ml into the nuclei using a custom setup microinjector and Olympus microscope (Olympus IMT-2) and the cells were returned to incubation at 37°C overnight to allow for protein expression. For DIC and fluorescence time-lapse analysis, cells were incubated on a heated stage at 37°C and imaged with a monochromator on a Zeiss Axiovert 200 M microscope enclosed in an incubator with CoolSNAP CCD camera.